Update README.md

README.md

@@ -1,13 +1,13 @@
 # COVID19-PRM
 The COVID19 Political Realities Model (PRM)
 
-The Political Realities Model (PRM) is built on observations of macro-level societal and political responses to COVID measured only in terms of infections and deaths. The starting point of the model is the belief that individuals' and policy makers' perceptions of COVID-19 risks generate a relatively narrow and fairly predictable, but not ideal, set of responses. Although specific individual behavior and government regulation may vary over time and from place to place, the range of what a society and its elected officials will accept as the aggregate impact from COVID19 can relatively easily be characterized. This model seeks to characterize these responses and models their impact on infection rates. Infection rates are in turn is related to deaths via a simple (time- and geographically-varying) phenomenological model derived through fitting. This allow the model to build in variation in the demographics of the infected population, access to health care & COVID testing and the actual infection fatality rate from place to place without explicitly considering these factors. At present the model is deterministic rather than probabilistic, and so only provides point forecasts. The model considers the United States at the state level.
+The Political Realities Model (PRM) is built on observations of macro-level societal and political responses to COVID19 characterized only in terms of infections and deaths. The starting point of the model is the hypothesis that although individuals and policy makers have responded to the epidemic with a wide variety of behavioral modifications and policy actions, the actual *net impact* of the measures taken, though not identical across time or geography, is predictable. Moreover, these predictions can be based upon what a society and its elected officials have demonstrated it (they) will accept as a reasonable aggregate impact from COVID19. This model seeks to identify one or more "acceptability" range(s) from observation of the epidemic up to the current time and to understand how much impact on infection rates is achieved by responses in this (these) range(s). Deaths, in turn, are related to infections via a simple (time- and geographically-varying) phenomenological model derived through fitting. This allows the model to build in spatiotemporal variation in the demographics of the population, its access to health care & COVID testing, and the actual infection fatality rate without explicitly considering these factors. At present the model is deterministic rather than probabilistic, and so only provides point forecasts. The model considers the United States at the state level.
 
-The phenomenological relationship between infections and deaths is, in the present version, a convolution of confirmed infections with a normal distribution death curve centered at 20 days delay from reported infection and having a sigma of 3 days, weighted by a time- and geographically-carying case fatality rate (CFR) factor derived from fitting (recent) past data. State-level CFRs are found to generally fall in the range of 0.015 to 0.025 as late July 2020. This approach could be modified by introducing flexibility in the shape of the death distribution curve (e.g., delay, width, skew) but tests show that this does not seem to be as important of a factor in the output of the model as the behavioral and policy choices (i.e., the "political choices.")
+The phenomenological relationship between infections and deaths is, in the present version, a convolution of confirmed infection counts with a normal distribution death curve centered at 20 days delay from reported infection and having a sigma of 3 days, weighted by a time- and geographically-varying case fatality rate (CFR) factor derived from fitting (recent) past data. State-level CFRs are found to generally fall in the range of 0.015 to 0.025 as late July 2020. This approach could be modified by introducing additional flexibility in the shape of the death distribution curve (e.g., delay, width, skew) but tests show that this does not seem to be as important of a factor in the output of the model as the behavioral and policy choices (i.e., the "political realities.")
 
-The "political choices" are modeled using a 3-level categorization of (7-day averaged) new infections per capita (low, medium and high). This categorization in turn affects changes in new infection rates in the model going forward (forcing down or allowing to go up), with limits based on observations from prior data. The observed daily rate in change of the (7-day averaged) new infections is rarely below 0.98 at the current time, but there is considerably more headroom above 1.0.
+The state of the epidemic (at present and future times) in each geography is described using a 3-level categorization of (7-day averaged) new infections per capita (low, medium and high). This categorization in turn affects changes in future infection rates in the model going forward (forcing down or allowing to go up), within limits based on observations from prior data. The observed daily rate in change of the (7-day averaged) new infections is rarely below 0.98 at the current time, but there is considerably more headroom above 1.0.
 
-This represents a very different approach from most COVID19 models, which generally attempt to model the epidemic with a first-principles approach (e.g., SEIR) and attempting to incorporate the impact of mulitple specific interventions (e.g., social distancing, school and business closures, reduced travel, use of facemasks) on the spread of disease. This model eschews that level of detail and instead relies on observations of the impacts of the effects of the variety of measures in use at a macro level. 
+This represents a very different approach from most COVID19 models, which generally attempt to model the epidemic with a first-principles approach (e.g., SEIR) and incorporate estimates of the impact of mulitple specific interventions (e.g., social distancing, school and business closures, reduced travel, use of facemasks) on the spread of disease. This model eschews that level of detail and instead relies on observations of the effects of the variety of measures in use at a macro level. 
 
 This is very much a work in progress. New developments will be described here.